Open-source Agent Framework OpenClaw Has High-risk Command Injection Vulnerability; Research Team Initiates Fix Process

A joint team from China Academy of Information and Communications Technology, Shanghai Jiao Tong University, and Nanjing University conducted a security audit on the open-source agent framework OpenClaw, discovering a high-risk LLM-driven command injection vulnerability. The flaw lies in the bash-tools module's lack of strict escaping for LLM-generated command parameters, allowing attackers to bypass defenses via manipulative prompts and execute remote code or exfiltrate sensitive data.

The team has initiated responsible disclosure, reporting findings and fixes to the Ministry of Industry and Information Technology's NVDB platform, and submitting GitHub issues to assist repairs. China Academy of Information and Communications Technology will continue monitoring OpenClaw's security risks to support safe industrial applications.